Sentiment Analysis of Weibo Text Using a Deep Combination Model
Jingzhong Li, Feng Chen · 2023
Social media platforms like Weibo have become the primary means for the public to express emotions and opinions. However, sentiment analysis with manual interpretation is not applicable in practice, due mainly to the vast amount of user-generated contents. This paper adopted an approach for text sentiment analysis, called CNN-BiLSTM model, through combining Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM). The CNN-BiLSTM model has been proved more effective in accurately determining the sentiment orientation within complex Weibo comments under various contexts. By employing the CNN-BiLSTM model, this research aimed to enhance the performance of sentiment analysis tasks, leading to improved accuracy and efficiency. Results demonstrated remarkable success, with the CNN-BiLSTM model achieving an accuracy of 98.16%, which suggested the efficacy of deep learning techniques in addressing the challenges of sentiment analysis for Weibo text. However, it is important to acknowledge some limitations in this study, such as the oversimplified sentiment labels, the relatively limited dataset scope, and the absence of consideration for multiple emotional nuances. To overcome these shortcomings, future endeavors should focus on introducing more nuanced sentiment labels, expanding the dataset's coverage, and accounting for additional factors influencing sentiment attitudes. Furthermore, more advanced and reliable deep learning models are required for sentiment attitude prediction.